Ronald R. Mourant and Beverly K. Jaeger. Virtual Environments Laboratory .... Driver view-port controlling module. ⢠Highway traffic simulation module.
Dynamic Evaluation of Pre-Entry HOV Signage Using a Virtual Environments Driving Simulator Ronald R. Mourant and Beverly K. Jaeger Virtual Environments Laboratory 334 Snell Engineering Center Northeastern University Boston, MA 02115 U.S.A. 1. INTRODUCTION The purpose of this study was to investigate the use of signage with respect to motorists using high-occupancy vehicle (HOV) lanes. The Association of State Highway and Transportation Officials has published a "Guide for the Design of High Occupancy Vehicle Facilities" (1). This guide recommends that HOV signs and lane markings conform to the Manual on Uniform Control Devices (MUTCD) (2) to the fullest extend possible, and that the diamond symbol (white on a black background) be incorporated into HOV sign format. In general, the present HOV regulatory signs follow the generic signage guidelines given in the MUTCD and have the following properties: 1) black letters on a white background, 2) rectangular shape, and 3) either reflectorized or illuminated during periods of low visibility. Most, but not all of the HOV signs have incorporated the diamond symbol. The recommend location for the diamond is the upper-left of a rectangular sign but this is not always adhered to. The recommendation that the diamond symbol be white on a black background is in conflict with the MUTCD guideline of black letters on a white background. The rational for the triangle being white may be that a white triangle symbol is used on pavement to designate an HOV lane. Some HOV signs have extended the black background for the triangle symbol across the width of the HOV sign and display additional information using white letters on a black background. Other HOV signs simply have the black background for the white triangle on a small portion of the sign. The result is that HOV regulatory signs are not presented in a consistent manner to motorists. It is stated in the report, "Preferential Lane Treatments for High-Occupancy Vehicles" (3), that the MUTCD does not provide clear guidelines with respect to HOV signage. There are concerns that HOV signs are not distinguishable by means of their background color. Motorists may tend to ignore them because they are just black and white and therefore considered to be of low priority. On the other hand, construction signs, with their orange background, are easily distinguishable to motorists who then realize that they may have to decrease their speed or change lanes, etc. Another concern expressed in (3) is that the diamond symbol is also being used for commercial truck lanes and bicycle lanes and routes. It has been suggested that the use and recommended application of the diamond symbol for HOV signage be periodically reevaluated.
In this study we examine HOV signage with respect to advance guidance for announcing an entrance to a HOV facility. When traveling South on Interstate 93 before reaching Boston there is a HOV lane. The Massachusetts Highway Department has noted a number of accidents just prior to the entrance of this facility. It appears that drivers make last minute decisions and/or erratic maneuvers as to the use of the HOV lane. This behavior may be responsible for rear-end and angle accidents when changing lanes to either take or not take the HOV lane. Our approach for studying HOV signage is to build a three-dimensional model of Interstate 93 beginning about three miles North of the HOV entrance and ending when the HOV lane terminates. This baseline model included the present HOV signs and lane markings as well as all other regulatory signs. Using this three-dimensional model, driver performance would then be measured as participants drove this section of Interstate 93 using Northeastern University's Virtual Environments Driving Simulator (4). Alternative HOV signing would then be designed. The alternative HOV signage would then replace the original HOV signs and measures of driving performance would be collected again. We would compare the performance under both conditions and make recommendations as to guidelines for HOV signs. Section 2 describes the Virtual Environments Driving Simulator. Our plan for data collection and instructions given to the study participants are presented in Section 3. Drivers' performance in terms of lane selection errors and time and distance to enter the correct lane with respect to a decision as to use or not use the HOV facility is shown in Section 4. In Section 5 recommendations for HOV signage are given. 2. VIRTUAL ENVIRONMENTS DRIVING SIMULATOR Northeastern University's Virtual Environments Driving Simulator (NUVEDS) consists of three modules: hardware, software, and the scenario database. 2.1 Hardware The hardware consisted of a real vehicle cab, a helmet mounted display, a head tracking device and a desktop computer. 2.1.1 Vehicle Cab A Dodge Caravan was purchased since the vehicle dynamics in the simulator are those of a Dodge Caravan. We modified the vehicle by removing the engine and cutting off the rear end. The remaining part of the vehicle included both front seats, the instrument panel, all driver controls, the windshield and the vehicle frame. The gas and brake pedals were instrumented with potentiometers and wired to the computer so the simulator software could calculate the acceleration and deceleration of the vehicle as a driver used the pedals. Similarly an optical encoded was installed on the steering column and the signal was transmitted to the computer to have the software calculate the direction of the vehicle. This real vehicle cab enabled participants to drive the simulated vehicle while using the controls of a real vehicle.
2.1.2 Helmet Mounted Display
The helmet mounted display had Dual 1.3” diagonal Active Matrix Liquid Crystal Displays which had a resolution per eye of : ((640x3)x480), (921,600 color elements) equivalent to 307,200 triads. The diagonal field of view was 60 degrees. 2.1.3 Head Tracking A gyro-based head tracking system was used with the new helmet mounted display. The IS-300 obtains its primary motion sensing using a miniature solid-state inertial measurement unit which senses angular rate of rotation, gravity and earth components along three perpendicular axes. The angular rates are integrated to obtain the orientation (yaw, pitch, and roll) of the sensor. 2.1.4 Computer The computer used to control NUVEDS was manufactured by Gateway and has the following features: Intel 400MHz Pentium II Processor with 512k Cache, and ATI Rage Pro Turbo 8MB 2x AGP Graphics Accelerator. This computer enabled frame rates to be between 15 and 20 frames/sec for the HOV test runs. 2.2 Software The software for the NUVEDS consists of the following modules: • • • • • • • • • • •
System initializing and controlling module 3-D highway model loading and controlling module Vehicle dynamics simulation module Vehicle motion controlling module Driver view-port controlling module Highway traffic simulation module Collision detection module Sound rendering module Image rendering module and Vehicle control device drivers Head movement tracking device drivers
2.3 Scenario Database In order to make the 3-D highway model accurately resemble the real highway environment and HOV facilities, we carried out a study of the segment of I-93 which contained the HOV lane. The data we collected in the study included: • • • • •
A television tape which showed the width, curvature, elevation, and number of lanes of the highway segment All signs and pavement markings (including HOV signs and pavement markings) The volume of traffic on the highway Landscapes Other highway structures, such as guard rails, bridges, etc.
Based on this data, a 3-mile long, 8-lane (4 lanes on each direction) highway model was built using the RenderWare 2.1 Scripting Language. In each direction four 12-foot lanes were modeled
with lane markings resembling those in the real world. Guard rail along the right most lane and Jersey barriers in the center median strip were accurately modeled. Particular attention was paid to the placement and design of the road signs. 3. EXPERIMENTAL METHOD In order to evaluate the HOV facility on I-93 and compare the original HOV signs with modified HOV signs, a experiment was carried out using the NUVEDS. Twenty NU engineering students participated in the experiment as subjects. Sixteen of them were males, and four were females. All of them had a minimum of one-year driving experience and were between the ages of 18 and 38. Subjects drove in a virtual highway environment (a highway segment modeled after I-93 southbound toward Boston) using our driving simulator. The HOV lane is open Monday-Friday from 6:30-9:30 a.m. to vehicles with 2 or more passengers. It does not permit exits to either Sullivan Square or Charlestown. The conditions that we studied were the original versus the modified traffic signs and whether a driver was eligible or ineligible to take the HOV lanes. This resulted in a 2 X 2 experimental design. Each subject made four runs, one under each combination. Thus data was collected for a total of 80 runs. For each run, a subject was given a set of instructions which applied to their trip: destination, number of occupants, and current time of the day ( It was a given condition that the subject was traveling on a weekday). In each run, a subject needed to make a decision as to whether to travel in the HOV lane or not. The dependent variables representing a driver’s performance included the following measures: • Lane change location in relation to sign position (distance before or distance after sign) • Velocity of the vehicle when approaching each sign • Number of errors in lane selection (if any) The subject was also instructed to travel in the HOV lane if eligible. Otherwise, the subject should not take the HOV lane. The subject then entered the vehicle, donned the HMD, and was given a practice session driving on a simulated roadway that resembles the highway segment used in the test protocol. The subject proceeded to the testing session once an adequate level of familiarity with the operation of the simulator was achieved. After the four runs, a questionnaire was administered to the subject to obtain feedback about the adequacy of the simulator, self-reported location of visual focus while driving, information contained on the signs, and level of assistance provided by the associated HOV signs. 4. RESULTS 4.1 Lane Selection Error In the eighty runs, seven lane selection errors occurred at the entrance to the HOV lane. Each error represents a wrong decision by a driver. Among the seven lane errors, five occurred in HOV1 runs, and two occurred in HOV2 runs. The two errors that occurred in HOV2 runs were both made by ineligible drivers who selected the HOV lane. Among the five errors that occurred in HOV1, two were made by eligible drivers who did not select the HOV lane, and the other three were made by ineligible drivers who selected the HOV lane. This data suggests that subjects made more accurate decision on HOV2 runs than on HOV1 runs.
4.2 Time and Distance Measures in HOV1 and HOV2 Runs In each run, we recorded the distance and the time between the decision point (i.e. the point where the subject decided to select or not select the HOV lane) and starting point of the run. The distance and time measures for HOV1 runs and HOV2 runs are shown in Table 4-1. Table 4-1 Distance and Time of HOV1 and HOV2 Runs HOV1 HOV2 HOV1 - HOV2 Difference Average Distance (m) 1059.3 744.13 315.3 +42.4% 3 Average Time (sec) 49.06 40.88 8.18 +20.0% The average distance for the subjects to make a decision in a HOV1 run was 315.3 m (42.4%) longer than that for the subjects to make a decision in a HOV2 run. The average time for subjects to make a decision in a HOV1 run is 8.18 sec (20.0%) longer than that for the subjects to make a decision in a HOV2 run. This means that some subjects made decisions much earlier in HOV2 runs than in HOV1 runs. Among the forty eligible runs, twenty of them are HOV1 runs, and the other twenty are HOV2 runs. The average distance and time between the decision point and the starting point of the eligible HOV1 and HOV2 runs are shown in Table 4-2. Table 4-2 Distance and Time of Eligible Runs HOV1 HOV2 Average Distance (m) Average Time (sec)
1241.7 6 54.25
Difference
663.62
HOV1 – HOV2 578.14
39.06
15.19
+38.8%
+91.6%
The average distance for the eligible subjects to finish the HOV1 runs was 578.14 m (91.6%) longer than that for the subjects to finish the HOV2 runs, and the average time for the eligible subjects to finish the HOV1 runs was 15.19 sec (38.8%) longer than that for the subjects to finish the HOV2 runs. It is obvious that the subjects made decisions quicker in HOV2 runs than in HOV1 runs. The average distance and time between decision the point and starting point of the ineligible HOV1 runs and HOV2 runs are shown in Table 4-3.
Table 4-3 Distance and Time of Ineligible Runs HOV1 HOV2 HOV1 - HOV2 Average Distance (m) 904.69 837.07 67.62
Difference +8.1%
Average Time (sec)
44.81
42.59
2.22
+5.2%
The average distance for the ineligible subjects to finish the HOV1 runs was 67.62 m (8.1%) longer than that for the subjects to finish the HOV2 runs. The average time for the ineligible subjects to finish the HOV1 runs was 2.22 sec (5.2%) longer than that for the subjects to finish the HOV2 runs. Although large differences were found for the eligible runs, only small differences were found for the ineligible runs. 4.3 t-Tests on the Data In order to validate the phenomena observed in the previous section, we did t-tests. Since the distance measures of the decision point and the time measures of decision point are correlated in our experiment, we only did t-tests on the distance measures (Table 4-4). Table 4-4 Calculations of t Values and Significance Levels HOV1 run vs. HOV2 run N
η s t significance
20 414.05 458.28 4.04 < 0.005
Eligible HOV1 run vs. Eligible HOV2 run 20 738.95 633.97 5.21 < 0.005
Ineligible HOV1 run vs. Ineligible HOV2 run 20 89.15 614.56 0.65 0.26
Based on the significance levels, we can conclude: • The modified HOV signs are overall more efficient in guidance quality than the original signs. • The modified HOV signs are especially effective in helping drivers that are eligible to use HOV lane. •
In term of helping drivers that are ineligible to use HOV lane, the modified HOV signs and the original HOV signs have about the same level of effectiveness.
4.4 Average Velocity of the Whole Run The average velocities of the whole run of HOV1 and HOV2 runs are shown in Table 5-5. The average velocity of HOV1 runs is 72.91 km/h, while that of HOV2 runs is 71.67 km/h. The difference between them is only 1.76%. It suggests that the modified HOV signs, although having one more sign than the original HOV signs, did not slow down the velocity of vehicles.
Table 4-5 Average Velocity of the Whole Run Distance (m) Average Time (sec)
HOV1 Runs 2110 104.19
HOV2 Runs 2110 105.98
Average Velocity (km/h)
72.91
71.67
5. DISCUSSION AND RECOMMENDATIONS In a study of vision and driving performance, Higgins, Wood, & Tate (5) reported that under conditions of optimal vision (20/20) subjects were not able to report the contents of all the road signs that they encountered. Therefore, essential information must be repeated on subsequent signage to account for division of attention between driving tasks and assessment of sign information. This is accomplished in 7 signs. Post-test questionnaire comments for the current study revealed that some subjects thought that there were too many signs presented in the highway segment (HOV1 contained 7, HOV2 contained 8). 5.1 Standardization: size, color, and symbols. All informational signs should be of the same rectangular shape, approximately 1:1.35 height-to-width ratio. All signs relating to the HOV lane should contain the diamond express symbol in the same upper-left location on the sign (also assists identification). All signs relating to the HOV lane should have the standard color pattern of black background on top with white lettering and white background on the bottom with black lettering (also assists identification). 5.2 Identification and Readability. A combination of both dark and light border frames should be used on each sign to improve contrast of sign contour against a variety of roadside backgrounds. Clearview font is recommended to enhance daytime and nighttime word legibility and recognition. This font resembles Arial typeface with increased interior spaces of the letters. Clearview font is designed to minimize the blurring effects of contrast irradiation to improve readability (6). The largest possible mixed-case lettering should be used to improve both the legibility of words and recognition of familiar terms on the highway signs (6). Upper-case words are used only for calling attention to the inaccessible destinations in signs 3 and 6. These exceptions are also underlined. 5.4 Content Post-test evaluations queried subjects regarding the maximum pieces of relevant information to be included on a sign. The number of manageable items of information were reported as follows: 50% percent reported 3, 45% reported 2, and 5% stated that 4 pieces of sign information could be readily processed while driving along the highway. Number of passengers, eligibility days and times, and destinations are considered to be primary information elements for the HOV selection task. Lane location and inclusion of buses are considered to be secondary elements of information. A sign shall contain a maximum of 2 pieces of primary information and one secondary piece when presented for the first time. No sign contains a destination (i.e., Boston) in conjunction with distance to lane. This avoids confounding the mileage to destination with the distance to lane entrance. 6. REFERENCES 1. "Guide for the Design of High Occupancy Vehicle Facilities". American Association of State Highway and Transportation Officials, 1992. 2. Manual on Uniform Traffic Control Devices. Federal Highway Administration, U.S.
Department of Transportation, 1988. 3. "Preferential Lane Treatments for High-Occupancy Vehicles". NCHRP Synthesis 185, Transportation Research Board, National Academy Press, 1993. 4. "A Driving Simulator Based on Virtual Environments Technology". Oren Levine and Ronald R. Mourant, Paper No. 950269, 74th Annual Meeting of the Transportation Research Board, January, 1995. 5. "Vision and Driving: Selective Effect of Optical Blur on Different Driving Tasks". K.E. Higgins, J. Wood and A.Tate, Human Factors, 40:2, June, 1998, pages 224-232. 6. "Clearer Road Signs Ahead". Garvey, P.M., Pietrucha, M.T., & Meeker, D.T.,. Ergonomics in Design, 6:3, July, 1998, pages 7-11.